Rapid Analysis Of Complex Systems By Ultraviolet Spectroscopy Combined With Chemometrics

Modern analytical chemistry is characterized by the instrumentalization of analytical methods and the complexity of analytical objects,wit h the goal of less consumables,low cost and high efficiency.Due to the exposure of water pollution and food adulteration,two typical complex systems,water environment and liquid food,have attracted much attention.The traditional analysis method is de structive,its speed is slow and even causes secondary pollution.In order to achieve on-line monitoring of water quality and on-site inspection of counterfeit food,making an efficient nondestructive analysis technology suitable for highly frequent repeat ed measurement,is a task which brooks no delay.In this paper,the environmental water samples and liquor samples taken as the research objects,the quantitative and qualitative modeling along with model optimization are discussed respectively by ultraviolet spectroscopy combined with chemometrics.The main results are as follows:In the aspect of quantitative analysis,the multivariate calibration based on UV spectra is studied to rapidly estimate concentrations of multicomponent in water.The effects of different spectral pretreatment methods were contrasted.Principal component regression(PCR),partial least squares regression(PLSR)and support vector regression(SVR)are used to construct calibration models of full band spectra.Competitive adaptive reweighted sampling(CARS),uninformative variable elimination(UVE),moving windows partial least squares(MWPLS),variable combination population analysis(VCPA)and other feature selection strategies are used to optimize the model,then compare the predicted results of different models,so as to establish the best calibration model of each component in water.The results have proven that ultraviolet spectroscopy combined with multivariate calibration is an efficient nondestructive method for assessing water quality.In the aspect of qualitative analysis,the pattern recognition based on UV spectra is mainly studied for the rapid classification of liquor species.The ultraviolet spectra data of liquor samples consisted of different geographical origins are collected.On the basis of appropriate spectral pretreatment,the quality control chart is constructed by calculating the angle cosine values,correlation coefficients and UV similarities between spectra of samples and the reference spectrum.Based on PCA-SVM and PLS-DA,the discriminant analysis model is established.The results have demonstrated that all of the above methods can realize the rapid identification of ?Huzhu‘ brand Qingke liquor.